Roadmap

Where Sigilweaver Loom stands today and what stands between it and a 1.0 release.

This page does not promise dates. Items move as the project learns what matters. The CHANGELOG is the record of what has shipped.

Where things stand

Loom is pre-1.0. There has been no stable release yet, so expect breaking changes between versions and some rough edges. The 0.13 line is meant to be a solid base: consistent tooling, security hardening, honest documentation and a repeatable release process. Feature work builds on it.

Component	Status	Notes
Studio	Beta	13 tools, tabs, undo/redo, themes, data preview and profiling, optional AI assistant.
Server	Beta	Lazy Polars execution, disk caching of blocking tools, Python code export, Hub integration.
Desktop apps	Beta	Releases ship Electron builds in a Full (bundled Server) and a Client variant, plus a web build. Tauri builds exist but are not part of the release yet. Windows builds are code signed. macOS builds are not notarized and target x64.
Docker deployment	Beta	Compose files for Hub, Server and PostgreSQL with secret handling. Images are amd64 only.
Hub	Open beta	Users, roles, groups, shared connections, scheduling, audit log. See Hub Open Beta for the current limitations.

Beta here means the features work and are tested, but file formats, configuration keys and APIs may still change, and there are known gaps.

The path to 1.0

Loom reaches 1.0 when these are true:

1. Stable contracts. The .swwf workflow format, the Server API and the Hub /api/v1 API have compatibility rules, and breaking changes come with a documented migration.
2. A security posture we can state plainly. Expression evaluation, file access, connection credentials and Hub tokens have been reviewed, and the Security Model describes exactly what is and is not guaranteed.
3. Hub is dependable for teams. Executions and schedules survive restarts, and Hub can run more than one replica.
4. Releases are boring. Every platform build is produced by CI, checksummed and installable without workarounds.
5. The docs match the code. Every tool, setting and environment variable is documented and checked before each release.

Focus areas

These are directions, roughly grouped. They are not ordered and not scheduled.

Studio

• Theme editor (a GUI for creating and editing themes)
• Workflow templates for common tasks
• Tool search in the palette
• Incremental preview for very large datasets
• UI for Hub features that exist only in the API today, such as workflow revisions and execution artifacts

Server

• More native database connectors
• More tools. Ideas include fuzzy join, network clustering and a Python code tool. None of these has been designed yet.
• Cloud storage sources (S3, GCS, Azure Blob)
• Native ARM64 builds

Hub

Hub is in open beta. The gaps that matter most:

• Durable execution runners. Records survive in PostgreSQL and stale runs are marked failed after restart, but Hub does not resume or retry the Server request.
• Supported multi-replica deployment. Schedule dispatch already uses transactional database claims; live execution events still need a shared broker, and the default in-memory token store must be replaced.
• A working Settings page in the admin area (it is a placeholder today)
• Per-user API tokens for automation
• OIDC and single sign-on
• Notifications (email and webhooks) for execution status

Platform and release

• Notarized macOS builds
• Helm charts and a Kubernetes deployment guide
• Broader testing on macOS and on Linux with X11

Already shipped

Some items that earlier versions of this page listed as future work are in place:

Area	Feature
Hub	Audit log browser with hash-chain verification (Admin > Audit Log)
Hub	Workflow version history in the web app: list, diff, download and restore
Hub	Execution artifacts: outputs stored as CSV and downloadable through the API (API only, no UI)
Hub	Live execution updates over server-sent events
Hub	Web-based management of servers, connections, users, groups and grant groups
Studio	AI assistant with several model providers
Studio	Data profiling and Python code export

Long-term principles

• Local first, cloud capable. Loom runs wherever your data lives, including air-gapped networks.
• Code as truth. A visual workflow can be exported as readable Polars code, so nothing is locked in.
• Open. The whole suite is AGPL-3.0, with no paid tiers.

Influence the roadmap

• Open or upvote a GitHub issue.
• Join a discussion about a larger change before you build it.
• Send a pull request. See the Contributing Workflow.